An open-source, terminal-based coding agent that works with many AI models. Bring your own API key, pick your model per session, and keep full control.
OpenCode is the hacker's choice: a terminal agent you can read, fork, and extend. Because it's model-agnostic, you're never locked into one provider — use a frontier model for hard tasks and a cheaper one for routine edits.
It does the full agent loop: explore the codebase, plan, edit files, run commands, and iterate on failures. The open-source community moves fast, so features land quickly.
Plug in API keys from major providers and switch models per task.
Inspect, fork, and customize the agent itself.
Lives where developers already work: shell, git, scripts.
Route simple tasks to cheap models, hard ones to frontier models.
Developers who want maximum control: pick your models, control costs, and aren't afraid of a terminal or an API key.
Pricing and features change fast — check the official site for the latest.
Visit official site →Last updated: October 2026 · Compare all agents